| Benchmark | DeepSeek-V3.2 | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 39.8 | 29 |
| AA Intelligence | 32.8 | 40.5 |
| AA-LCR | 70.7 | 72.3 |
| AA-Omniscience | -22.5 | 2.6 |
| aime_2026 | 95.1 | 95.3 |
| arena_text_factuality | 1433 | 1438 |
| Artificial Analysis Coding Index | 44.2 | 54.5 |
| coding_arena_elo | 1368 | 1460 |
| critpt | 2.9 | 2.9 |
| frontiermath_tier_4 | 2.1 | 8.3 |
| gdpval | 18.8 | 32 |
| GPQA Diamond | 84 | 90.4 |
| HLE | 40.8 | 28.8 |
| HLE (with tools) | 40.8 | 50.6 |
| HMMT 2025 | 90.2 | 94.6 |
| hmmt_feb_2026 | 84.1 | 87.8 |
| hmmt_nov_2025 | 90.2 | 94.6 |
| IFBench | 61 | 75.2 |
| imo_answer_bench | 78.3 | 83.8 |
| LiveCodeBench v6 | 83.3 | 87.1 |
| longbench_v2 | 59.8 | 62 |
| MCP Atlas | 62.2 | 74.1 |
| mcpmark | 38 | 48.2 |
| mmlu_redux | 93.7 | 94.5 |
| MMLU-Pro | 86 | 88.5 |
| OmniScience Accuracy | 33 | 26.4 |
| OmniScience Non-Hallucination | 17.3 | 68 |
| scicode | 39 | 40.7 |
| simpleqa_verified | 27.5 | 49.1 |
| SWE-bench Multilingual | 70.2 | 73.8 |
| SWE-bench Pro | 15.6 | 56.6 |
| SWE-bench Verified | 73.1 | 78.8 |
| TauBench V3 - Banking | 18.8 | 20.8 |
| Terminal-Bench 2.0 | 46.4 | 61.6 |
| Terminal-Bench 2.1 | 46.8 | 61.4 |
| Terminal-Bench Hard | 35.6 | 43.9 |
| tool_decathlon | 35.2 | 39.8 |
| toolathlon | 35.2 | 39.8 |
| τ²-Bench Telecom (AA run) | 90.6 | 97.7 |
| τ³-Bench | 69.2 | 70.7 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.